1 |
/***************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - SAD calculation module (C part) - |
* - Sum Of Absolute Difference related code - |
5 |
* |
* |
6 |
* Copyright(C) 2002 Michael Militzer <isibaar@xvid.org> |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
|
* 2002 Peter Ross <pross@xvid.org> |
|
|
* |
|
|
* This program is an implementation of a part of one or more MPEG-4 |
|
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
|
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
7 |
* |
* |
8 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
9 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
|
* $Id$ |
23 |
|
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
25 |
|
|
26 |
#include "../portab.h" |
#include "../portab.h" |
27 |
|
#include "../global.h" |
28 |
#include "sad.h" |
#include "sad.h" |
29 |
|
|
30 |
|
#include <stdlib.h> |
31 |
|
|
32 |
sad16FuncPtr sad16; |
sad16FuncPtr sad16; |
33 |
sad8FuncPtr sad8; |
sad8FuncPtr sad8; |
34 |
sad16biFuncPtr sad16bi; |
sad16biFuncPtr sad16bi; |
35 |
sad8biFuncPtr sad8bi; // not really sad16, but no difference in prototype |
sad8biFuncPtr sad8bi; /* not really sad16, but no difference in prototype */ |
36 |
dev16FuncPtr dev16; |
dev16FuncPtr dev16; |
37 |
|
sad16vFuncPtr sad16v; |
38 |
|
|
39 |
sadInitFuncPtr sadInit; |
sadInitFuncPtr sadInit; |
40 |
|
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
|
|
|
|
#define MRSAD16_CORRFACTOR 8 |
|
|
uint32_t |
|
|
mrsad16_c(const uint8_t * const cur, |
|
|
const uint8_t * const ref, |
|
|
const uint32_t stride, |
|
|
const uint32_t best_sad) |
|
|
{ |
|
|
|
|
|
uint32_t sad = 0; |
|
|
int32_t mean = 0; |
|
|
uint32_t i, j; |
|
|
uint8_t const *ptr_cur = cur; |
|
|
uint8_t const *ptr_ref = ref; |
|
|
|
|
|
for (j = 0; j < 16; j++) { |
|
|
for (i = 0; i < 16; i++) { |
|
|
mean += ((int) *(ptr_cur + i) - (int) *(ptr_ref + i)); |
|
|
} |
|
|
ptr_cur += stride; |
|
|
ptr_ref += stride; |
|
|
|
|
|
} |
|
|
mean /= 256; |
|
|
|
|
|
for (j = 0; j < 16; j++) { |
|
|
|
|
|
ptr_cur -= stride; |
|
|
ptr_ref -= stride; |
|
|
|
|
|
for (i = 0; i < 16; i++) { |
|
|
|
|
|
sad += ABS(*(ptr_cur + i) - *(ptr_ref + i) - mean); |
|
|
if (sad >= best_sad) { |
|
|
return MRSAD16_CORRFACTOR * sad; |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
return MRSAD16_CORRFACTOR * sad; |
|
|
|
|
|
} |
|
|
|
|
41 |
|
|
42 |
uint32_t |
uint32_t |
43 |
sad16_c(const uint8_t * const cur, |
sad16_c(const uint8_t * const cur, |
47 |
{ |
{ |
48 |
|
|
49 |
uint32_t sad = 0; |
uint32_t sad = 0; |
50 |
uint32_t i, j; |
uint32_t j; |
51 |
uint8_t const *ptr_cur = cur; |
uint8_t const *ptr_cur = cur; |
52 |
uint8_t const *ptr_ref = ref; |
uint8_t const *ptr_ref = ref; |
53 |
|
|
54 |
for (j = 0; j < 16; j++) { |
for (j = 0; j < 16; j++) { |
55 |
|
sad += abs(ptr_cur[0] - ptr_ref[0]); |
56 |
|
sad += abs(ptr_cur[1] - ptr_ref[1]); |
57 |
|
sad += abs(ptr_cur[2] - ptr_ref[2]); |
58 |
|
sad += abs(ptr_cur[3] - ptr_ref[3]); |
59 |
|
sad += abs(ptr_cur[4] - ptr_ref[4]); |
60 |
|
sad += abs(ptr_cur[5] - ptr_ref[5]); |
61 |
|
sad += abs(ptr_cur[6] - ptr_ref[6]); |
62 |
|
sad += abs(ptr_cur[7] - ptr_ref[7]); |
63 |
|
sad += abs(ptr_cur[8] - ptr_ref[8]); |
64 |
|
sad += abs(ptr_cur[9] - ptr_ref[9]); |
65 |
|
sad += abs(ptr_cur[10] - ptr_ref[10]); |
66 |
|
sad += abs(ptr_cur[11] - ptr_ref[11]); |
67 |
|
sad += abs(ptr_cur[12] - ptr_ref[12]); |
68 |
|
sad += abs(ptr_cur[13] - ptr_ref[13]); |
69 |
|
sad += abs(ptr_cur[14] - ptr_ref[14]); |
70 |
|
sad += abs(ptr_cur[15] - ptr_ref[15]); |
71 |
|
|
72 |
for (i = 0; i < 16; i++) { |
if (sad >= best_sad) |
|
|
|
|
sad += ABS(*(ptr_cur + i) - *(ptr_ref + i)); |
|
|
|
|
|
if (sad >= best_sad) { |
|
73 |
return sad; |
return sad; |
|
} |
|
|
|
|
|
|
|
|
} |
|
74 |
|
|
75 |
ptr_cur += stride; |
ptr_cur += stride; |
76 |
ptr_ref += stride; |
ptr_ref += stride; |
81 |
|
|
82 |
} |
} |
83 |
|
|
|
|
|
|
|
|
84 |
uint32_t |
uint32_t |
85 |
sad16bi_c(const uint8_t * const cur, |
sad16bi_c(const uint8_t * const cur, |
86 |
const uint8_t * const ref1, |
const uint8_t * const ref1, |
98 |
|
|
99 |
for (i = 0; i < 16; i++) { |
for (i = 0; i < 16; i++) { |
100 |
int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; |
int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; |
101 |
|
sad += abs(ptr_cur[i] - pixel); |
|
if (pixel < 0) { |
|
|
pixel = 0; |
|
|
} else if (pixel > 255) { |
|
|
pixel = 255; |
|
|
} |
|
|
|
|
|
sad += ABS(ptr_cur[i] - pixel); |
|
102 |
} |
} |
103 |
|
|
104 |
ptr_cur += stride; |
ptr_cur += stride; |
128 |
|
|
129 |
for (i = 0; i < 8; i++) { |
for (i = 0; i < 8; i++) { |
130 |
int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; |
int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; |
131 |
|
sad += abs(ptr_cur[i] - pixel); |
|
if (pixel < 0) { |
|
|
pixel = 0; |
|
|
} else if (pixel > 255) { |
|
|
pixel = 255; |
|
|
} |
|
|
|
|
|
sad += ABS(ptr_cur[i] - pixel); |
|
132 |
} |
} |
133 |
|
|
134 |
ptr_cur += stride; |
ptr_cur += stride; |
149 |
const uint32_t stride) |
const uint32_t stride) |
150 |
{ |
{ |
151 |
uint32_t sad = 0; |
uint32_t sad = 0; |
152 |
uint32_t i, j; |
uint32_t j; |
153 |
uint8_t const *ptr_cur = cur; |
uint8_t const *ptr_cur = cur; |
154 |
uint8_t const *ptr_ref = ref; |
uint8_t const *ptr_ref = ref; |
155 |
|
|
156 |
for (j = 0; j < 8; j++) { |
for (j = 0; j < 8; j++) { |
157 |
|
|
158 |
for (i = 0; i < 8; i++) { |
sad += abs(ptr_cur[0] - ptr_ref[0]); |
159 |
sad += ABS(*(ptr_cur + i) - *(ptr_ref + i)); |
sad += abs(ptr_cur[1] - ptr_ref[1]); |
160 |
} |
sad += abs(ptr_cur[2] - ptr_ref[2]); |
161 |
|
sad += abs(ptr_cur[3] - ptr_ref[3]); |
162 |
|
sad += abs(ptr_cur[4] - ptr_ref[4]); |
163 |
|
sad += abs(ptr_cur[5] - ptr_ref[5]); |
164 |
|
sad += abs(ptr_cur[6] - ptr_ref[6]); |
165 |
|
sad += abs(ptr_cur[7] - ptr_ref[7]); |
166 |
|
|
167 |
ptr_cur += stride; |
ptr_cur += stride; |
168 |
ptr_ref += stride; |
ptr_ref += stride; |
173 |
} |
} |
174 |
|
|
175 |
|
|
|
|
|
|
|
|
176 |
/* average deviation from mean */ |
/* average deviation from mean */ |
177 |
|
|
178 |
uint32_t |
uint32_t |
200 |
for (j = 0; j < 16; j++) { |
for (j = 0; j < 16; j++) { |
201 |
|
|
202 |
for (i = 0; i < 16; i++) |
for (i = 0; i < 16; i++) |
203 |
dev += ABS(*(ptr_cur + i) - (int32_t) mean); |
dev += abs(*(ptr_cur + i) - (int32_t) mean); |
204 |
|
|
205 |
ptr_cur += stride; |
ptr_cur += stride; |
206 |
|
|
208 |
|
|
209 |
return dev; |
return dev; |
210 |
} |
} |
211 |
|
|
212 |
|
uint32_t sad16v_c(const uint8_t * const cur, |
213 |
|
const uint8_t * const ref, |
214 |
|
const uint32_t stride, |
215 |
|
int32_t *sad) |
216 |
|
{ |
217 |
|
sad[0] = sad8(cur, ref, stride); |
218 |
|
sad[1] = sad8(cur + 8, ref + 8, stride); |
219 |
|
sad[2] = sad8(cur + 8*stride, ref + 8*stride, stride); |
220 |
|
sad[3] = sad8(cur + 8*stride + 8, ref + 8*stride + 8, stride); |
221 |
|
|
222 |
|
return sad[0]+sad[1]+sad[2]+sad[3]; |
223 |
|
} |
224 |
|
|
225 |
|
uint32_t sad32v_c(const uint8_t * const cur, |
226 |
|
const uint8_t * const ref, |
227 |
|
const uint32_t stride, |
228 |
|
int32_t *sad) |
229 |
|
{ |
230 |
|
sad[0] = sad16(cur, ref, stride, 256*4096); |
231 |
|
sad[1] = sad16(cur + 8, ref + 8, stride, 256*4096); |
232 |
|
sad[2] = sad16(cur + 8*stride, ref + 8*stride, stride, 256*4096); |
233 |
|
sad[3] = sad16(cur + 8*stride + 8, ref + 8*stride + 8, stride, 256*4096); |
234 |
|
|
235 |
|
return sad[0]+sad[1]+sad[2]+sad[3]; |
236 |
|
} |
237 |
|
|
238 |
|
|
239 |
|
|
240 |
|
#define MRSAD16_CORRFACTOR 8 |
241 |
|
uint32_t |
242 |
|
mrsad16_c(const uint8_t * const cur, |
243 |
|
const uint8_t * const ref, |
244 |
|
const uint32_t stride, |
245 |
|
const uint32_t best_sad) |
246 |
|
{ |
247 |
|
|
248 |
|
uint32_t sad = 0; |
249 |
|
int32_t mean = 0; |
250 |
|
uint32_t i, j; |
251 |
|
uint8_t const *ptr_cur = cur; |
252 |
|
uint8_t const *ptr_ref = ref; |
253 |
|
|
254 |
|
for (j = 0; j < 16; j++) { |
255 |
|
for (i = 0; i < 16; i++) { |
256 |
|
mean += ((int) *(ptr_cur + i) - (int) *(ptr_ref + i)); |
257 |
|
} |
258 |
|
ptr_cur += stride; |
259 |
|
ptr_ref += stride; |
260 |
|
|
261 |
|
} |
262 |
|
mean /= 256; |
263 |
|
|
264 |
|
for (j = 0; j < 16; j++) { |
265 |
|
|
266 |
|
ptr_cur -= stride; |
267 |
|
ptr_ref -= stride; |
268 |
|
|
269 |
|
for (i = 0; i < 16; i++) { |
270 |
|
|
271 |
|
sad += abs(*(ptr_cur + i) - *(ptr_ref + i) - mean); |
272 |
|
if (sad >= best_sad) { |
273 |
|
return MRSAD16_CORRFACTOR * sad; |
274 |
|
} |
275 |
|
} |
276 |
|
} |
277 |
|
|
278 |
|
return MRSAD16_CORRFACTOR * sad; |
279 |
|
|
280 |
|
} |
281 |
|
|
282 |
|
|